Fermented foods: the microbes you eat aren't the point

Open a jar of live sauerkraut and you are opening an ecosystem. But what is in the jar and what happens inside you are two different questions, and only one of them has a randomised trial behind it.

What is actually alive in the jar

A fermented food is one made through desired microbial growth and enzymatic conversions of food components — the wording an expert panel convened by the International Scientific Association for Probiotics and Prebiotics settled on ². Notice what it does not say: nothing about the microbes still being alive when the food reaches your mouth. Some are. Many are not.

When they are, the numbers are large and inconsistent. A review of published counts in retail fermented foods by Rezac and colleagues at the University of Nebraska found yoghurt cultures from under 10⁴ to 10⁹ CFU per gram or mL ³; commercial kimchi at 7.14–9.23 log CFU/g of lactic acid bacteria ³; sauerkraut anywhere from 3.79 to 8.3 log CFU/g depending on the study, the top figure from a US product made with a starter culture ³; kefir made in Korea from commercial grains at 9.62 log CFU/g of lactic acid bacteria alongside 7.67 log CFU/g of yeast ³. At the top end, ISAPP puts up to 10¹¹ lactic acid bacteria in one serving of a live yoghurt or kefir ².

The four are not versions of one thing. Yoghurt is the narrowest: two organisms by definition, Lactobacillus delbrueckii subsp. bulgaricus — a Lactobacillus — and Streptococcus thermophilus — the combination the Codex Alimentarius requires ²⁵. Kefir is the widest: the Codex standard describes a fermented milk carrying Lentilactobacillus kefiri with Leuconostoc, Lactococcus and Acetobacter species, plus lactose-fermenting and non-lactose-fermenting yeasts ². Kimchi and sauerkraut are usually wild ferments — what grows is whatever was on the cabbage — though some commercial sauerkraut is made with a starter culture, and it is the sample that counted highest ³.

The trial that put a number on it

Wastyk, Fragiadakis and colleagues at Stanford ran a randomised study over a 17-week protocol — a 10-week diet intervention with three weeks of baseline before it and four weeks of follow-up after ¹: 36 generally healthy adults, 18 per arm ¹, on either a high-fibre or a high-fermented-food diet. A serving was defined as 6 oz of kombucha, yoghurt or kefir, a quarter-cup of kimchi or sauerkraut, 2 oz of vegetable brine drink ¹. A four-week ramp led to a target of six servings a day ¹, and the fermented-food arm averaged 6.3 ± 2.9 servings a day at the end of maintenance ¹.

Two things happened. Gut microbial alpha diversity rose on three measures — observed ASVs, PD whole tree and Shannon ¹ — and stayed up through the four-week choice period afterwards ¹. And 19 of 93 measured inflammatory serum proteins fell, IL‑6, IL‑10 and IL‑12b among them ¹.

The high-fibre arm did neither: diversity did not move cohort-wide ¹, and none of the 19 cytokines that fell in the fermented arm changed in the fibre arm ¹. And the caveat, out loud, because the paper says it: the study's primary outcome, a cytokine response score, was not significant for either arm ¹. The headline result here is a secondary one.

Why diversity rose, when the microbes did not stay

This is the finding that should change how you think about the jar. The Stanford team sequenced the foods as well as the stool, asking whether the new organisms in people's guts were ones they had eaten. Mostly they were not: overlap between new gut ASVs and fermented-food ASVs peaked at 5.4% early in the intervention and, later, was no different in the fermented arm than in the fibre arm ¹. The authors conclude the diversity increase was not primarily due to consumed microbes, but to shifts in, or new acquisitions to, the resident community ¹.

Mechanistically that is the more interesting story. A fermented food is not only cells; it is everything the cells made while the jar sat there — organic acids, peptides, broken-down sugars — arriving daily in a gut whose residents respond to that chemistry. Which of those does the work, nobody has isolated. What the trial rules out is the picture the marketing runs on: swallow bacteria, bacteria move in, diversity goes up.

What pasteurisation takes out

The live half, entirely. The ISAPP panel is explicit that heat-treated or pasteurised fermented vegetables, fermented sausage, soy sauce, vinegar and some kombuchas contain no living microorganisms by the time they are eaten ². Leavened bread is baked after fermentation, which kills the organisms outright; wine and most beers go through steps that remove the live microorganisms; coffee and chocolate beans are roasted ².

So shelf-stable sauerkraut and the refrigerated jar are not the same food for this purpose. Both were fermented; only one is alive. What fermentation built into the flavour survives the heat; the ecology does not.

The one benefit that really is the swallowed cells

One exception runs the other way. People who cannot digest lactose tolerate it in yoghurt, and the reason is the swallowed bacteria: their lactase is protected inside the cells, survives the stomach, and works in the small intestine as the pH rises and transit slows ⁵. Commercial yoghurts do this about equally well because they all carry roughly 10⁸ bacteria per mL ⁵. Europe's approved health claim for lactose digestion is written around live yoghurt cultures ²; heat the yoghurt and the cells go, and the lactase sheltered inside them goes too.

What a week looks like

Six servings a day is what was tested ¹, and it is a lot. It is also the only amount with a controlled trial behind it, so treat it as the tested dose, not the minimum useful one — the trial had no smaller-dose arm ¹.

A week that reaches it: a 6 oz pot of live yoghurt or kefir at breakfast ¹, a quarter-cup of kimchi or sauerkraut with lunch and with dinner ¹, and something fermented beside two other meals. Rotate rather than repeat: the four carry different organisms in different numbers ²³, and the trial fed a mixture, not a product.

The observational picture is more modest. Taylor and colleagues compared stool from 6,811 people in the American Gut Project ⁴ and found differences between consumers and non-consumers that were statistically significant but, in their own word, subtle ⁴, with an enrichment of conjugated linoleic acid among consumers ⁴. Real. Small. Not proof of cause.

What this does not tell you

It does not tell you fermented food treats anything: 36 healthy adults over a 17-week protocol ¹ is a mechanism study with a missed primary endpoint ¹. It does not tell you a capsule reproduces the effect; nobody knows which part of the food is responsible. And it does not tell you your jar holds what the study's jars held — the counts swing by orders of magnitude between brands and shelf lives ³.

What it does tell you is short enough to carry to the shop. Buy it cold, buy it live, eat it most days, vary what you eat. A dull instruction resting on an interesting finding: the bacteria you swallow matter less than what they leave behind.